Predictors of Adoption and Guardianship Dissolution: The Role of Race, Age, and Gender among Children in Foster Care

Abstract

Adoption and guardianship are meant to provide permanency to foster children when reunification is not a viable option. Unfortunately, sometimes adoption and guardianship placements dissolve resulting in children returning to care. Currently, there is limited research on the prevalence and predictors of adoption and guardianship dissolutions. This study investigated rate of guardianship and adoption dissolution using a complete entry cohort from a large state foster care system and the associations between child characteristics and risk factors with dissolution. Drawing on a complete entry cohort of foster children in Texas that exited to either adoption or guardianship placements, results demonstrated that over 2% of adoptive placements and 7% of guardianship placements were dissolved. Compared with White and Hispanic children, Black children had a higher risk of guardianship, but not adoption, dissolution. Older age was associated with a higher risk of adoption dissolution, and females had a higher risk of guardianship dissolution than males. Behavior problems, cognitive disability status, and mental health issues were all associated with a higher risk of dissolution. These findings have important implications for caseworkers and policymakers on permanency for children in adoptive or guardianship placements.

Nearly 700,000 children were served by the U.S. foster care system in 2017, almost all of whom have a case goal of permanency through reunification (returning to their birth families), adoption, or guardianship (or similar legal custody arrangement) (U.S. Department of Health and Human Services [USDHHS], 2018). Permanency is widely viewed as critical to promoting healthy development for foster children. Reunification is the most preferred form of permanency, but many families are not able to meet the criteria for safe reunification within the allotted time frame. Consequently, more than a third of children exiting foster care each year are adopted or placed in guardianship (USDHHS, 2018). Yet, adoption and guardianship are sometimes impermanent, and children return to foster care after an exit to permanency. Prior research has mostly focused on reentry to foster care following reunification (for review, see Kimberlin, Anthony, & Austin, 2009), with far less research on reentry following guardianship or adoption. This study examined the prevalence and demographic variation in adoption and guardianship dissolution using a longitudinal dataset from a large southern state.

Public Policies and Permanency

Policymakers have long expressed concern about the length of time children spend in foster care. A series of federal laws have attempted to reduce time in care and increase the number of prospective adoptive families for children. These efforts include financial incentives to states to move eligible children into adoptive families, subsidies for families who adopt special needs children, concurrently planning for reunification and alternative permanency arrangements, expediting termination of parental rights when reunification is unlikely to be successful, and oversight of timely permanency through federal Child and Family Services Reviews and mandatory data collection efforts (Adoption Assistance and Child Welfare Act 1980; Adoption and Safe Families Act 1997; Adoption Promotion Act of 2003). Regulations also sought to reduce time children spend waiting for permanency by limiting consideration of race in decisions about adoptive placements (Multiethnic Placement Act of 1994; Interethnic Adoption Act 1996). By the early 2000s, it was acknowledged that some children preferred not to be adopted and relatives were sometimes reluctant to pursue adoption (Geen, 2003). Guardianship, unlike adoption, does not require termination of parental rights. Thus, guardianship became more widely accepted and promoted as a permanency option, particularly for children living with relatives. Federal funds have been available to subsidize guardianships since 2008, in which financial support for guardianships is more comparable to adoptions (Fostering Connections to Success and Increasing Adoptions Act, 2008).

Permanency Dissolution

When children return to foster care after an adoption or a guardianship has been finalized, this is referred to as permanency dissolution (Rolock & White, 2017). The anticipated benefits associated with permanency –including stability and a sense of belongingness for children –may be negated if children return to foster care. The prevalence of adoption and guardianship dissolution is not well-tracked. National datasets may not accurately track individuals who reenter following adoption, particularly when adoptees have changed their names (Festinger & Maza, 2009). Existing research has typically used convenience or small samples, does not distinguish between adoption and guardianship, or focuses on adoption disruption (pre-finalization breakdown of a planned adoption). A handful of studies indicate adoption dissolution rates ranging from 1% to 5% (for review see Coakley & Berrick, 2008). Unfortunately, these studies have several limitations including small sample sizes, short follow-up periods, and samples from the 1970s-1990s. Guardianships are a more recent permanency option and the subject of far less research (White, 2016); however, reported dissolution rates range from 1% to 17% across studies (Goering & Shaw, 2017 [unnamed Northeastern State]; Koh & Testa, 2011 [Illinois]; Parolini et al., 2018 [California]; Rolock, 2015 [Illinois]; Rolock & White, 2017 [Illinois]). Given the range of dissolution rates and data limitations, the first aim was to identify the percentage of adoptive and guardianship placements that dissolve in a large southern state.

Although adoption and guardianship are both forms of permanency, children in these arrangements likely differ in their characteristics and risk of dissolution. Adoption requires the severance of legal ties between the child and their birth parents, which guarantees adoptive parents the same legal rights as birth parents. Guardianship does not require the termination of parental rights and can be dissolved by court order. The most common adoptive parents and guardians are children’s (kin or non-kin) foster parents. Adoptions may involve kin or non-relative caregivers and are more common among younger children, whereas guardianships tend to be more common among older children and children living with relatives (Koh & Testa, 2011). Although there is comparatively little research on kin adoptive parents and guardians, much can be gleaned from research on kinship foster parents. Compared to non-kin, kinship foster parents are typically older and more likely to be single, unemployed, and of lower socioeconomic status (Cuddeback, 2004) and their decisions to continue placement of a child appear to be affected by the size of financial supports (Doyle, 2007; Testa & Slack, 2002). Coupled with evidence that socioeconomic status is associated with permanency dissolution (Coakley & Berrick, 2008; Liao & White, 2014; Testa et al., 2015), this suggests guardianships and kin adoptions may be at higher risk of dissolution. Yet, kinship foster care is often more stable than non-kin foster care despite socioeconomic disadvantage (Winokur, Holtan, & Batchelder, 2014); suggesting that kin-based permanency (adoption or guardianship) may be more stable as well.

Beyond potential differences by kinship status, guardianships may be less stable than adoptions if caregivers see guardianship as less of a commitment than adoption. Denby (2011) found that kin caregivers with less emotional distress, caregivers who were given more detailed information about the case plan, and caregivers caring for children with fewer mental and behavior health problems were more likely to adopt, indicating that guardianships may involve more challenging children and less-prepared caregivers.

Race and Permanency Dissolution

Disparities persist in the timeliness and actualization of adoptions for racial and ethnic minority children, especially for Black children. Previously, disparities were attributed to delays due to efforts to find a racially matched placement (Shaw, 1998). However, for over two decades, federal policy has prohibited the use of the adoptive parent or child’s race, culture, national origin, or ethnicity to delay or deny an adoptive placement except in limited circumstances where it is in a child’s best interests (Interethnic Adoption Act 1996). Yet, Black and Hispanic children still wait longer to be adopted from foster care than White, Asian, or Pacific Islander children (Snowden, Leon, & Sieracki, 2008). In contrast, available data do not indicate consistent differences in selection into guardianship (Akin, 2011).

After adoption or guardianship, Black children may also face a higher rate of dissolution. In a study of over 51,000 Illinois youth who were adopted or placed in guardianship from 1998 to 2010, Black children were more likely to return to care than White children (Rolock & White, 2016), but adoption and guardianship dissolution were not examined separately. Studies in Colorado and Illinois found that Black children and (to a lesser extent) Hispanic children were more likely than White children to return to foster care after adoption (Orsi, 2015; Smith, Howard, Garnier, & Ryan, 2006). In California, Black children, but not Latino children, had a higher risk of foster care reentry after guardianship than White children (Parolini et al., 2018). Although these studies provide some evidence of racial disparities in permanency dissolution, each study either investigated only adoption or guardianship or combined both placement types.

There are several potential explanations for racial disparities in adoption and guardianship dissolution. First, transracial adoptions constitute about 17% of all adoptions from foster care (Hansen & Pollack, 2007). About 20% of adoptions of Black children and 38% of adoptions of Hispanic children were transracial compared with 3% of adoptions of White children (Hansen & Simon, 2004). In Texas (the site of the current study), approximately 42% of adoptions were transracial in 2019 (DFPS Data Book, 2020). Although children in transracial adoptions can develop safely and healthily with supportive and capable parents, transracial adoptions may require parents with additional skills or supports to be successful (Smith, McRoy, Freundlich, & Kroll, 2008). Under longstanding federal law, agencies cannot impose different requirements for a transracial placement than a same-race placement. Agencies can require cultural competency or other trainings to support transracial adoptions if required for all adoptions, not just transracial adoptions. Concerns about compliance with federal law may have led agencies to avoid such trainings entirely (U.S. Commission on Civil Rights, 2010). Without adequate preparation and support, transracial adoptions may be at higher risk of dissolution. In contrast, guardianships are typically with relatives and thus less likely to be transracial.

Second, race may affect how adoptive parents or guardians evaluate behavior problems. Studies have shown that teacher trainees view Black children, particularly Black males, as more hostile when exhibiting misbehavior than White children exhibiting similar behaviors (Halberstadt, Castro, Chu, Lozada, & Sims, 2018). There is also evidence of biases linking stereotypical Black features with dangerousness or risk among Black and non-Black observers (Kahn & Davies, 2011). These findings may extend to interactions between children and their adoptive or guardian caregivers, but this has not yet been evaluated. It is possible that, net of the occurrence of behavior problems and whether a placement is same-race or transracial, a child’s race may affect how adoptive parents or guardians respond to adjustment or behavior problems.

Third, there may be racial differences in children’s experiences prior to permanency—in foster care and with their families of origin-- that influence their adjustment. Racial disparities in childhood experiences – including exposure to child maltreatment, poverty, and other adversities may continue to affect children’s experience in and after foster care (Maguire-Jack, Lanier, & Lombardi, 2019). In addition, there is some indication that Black children have more difficult experiences in foster care. For example, Black children in Texas foster care experience higher rates of placement disruption than White and Hispanic children (Sattler, Font, & Gershoff, 2018). These experiences of multiple placement disruptions may undermine children’s ability to attach to and trust their adoptive parents or guardians (Dozier, Zeanah & Bernard, 2013), which may contribute to dissolution risk.

Overarching Factors Affecting Permanency Dissolution

Other social and demographic factors likely affect children’s post-permanency experiences. Older children are less likely to be adopted and more likely to age out of foster care (Annie E. Casey Foundation, 2018). If adopted or placed in guardianship, they face greater difficulty adjusting to the adoptive or guardianship arrangement and higher risk of dissolution (Coakley & Berrick, 2008; Orsi, 2015; Parolini et al., 2018; Smith et al., 2006). Children who enter foster care at older ages may have faced longer exposure to maladaptive environments and maltreatment. Consistent with such exposure, they have higher rates of social-emotional problems that can undermine permanency (McMillen et al., 2005). Moreover, because older children have spent a greater duration of time with their families of origin prior to entering foster care, they may have a more strong and sustained identification with their family of origin than younger children, which may make integration into a new permanent arrangement more challenging.

In addition, male children are less likely to be adopted than female children (Snowden et al., 2008), and there is a positive, but inconsistent, association between male gender and dissolution (Coakley & Berrick, 2008). Among children in foster care, males are more likely than females to have been removed partly due to behavior problems and to experience restrictive (congregate care) placements (U.S. Children’s Bureau, 2015). These differences may indicate a higher rate of behavioral problems among males or a tendency to view male children’s behaviors as more threatening than females (Hudley et al.,2001). Given the consistent association between caregiver-perceived behavior problems and the disruption of foster, adoptive, and guardianship placements (Coakley & Berrick, 2008; Liao, 2016; Parolini et al., 2018), we expect that older children and male children will have a higher risk of adoption or guardianship dissolution.

In addition to possible main associations of race, gender, and age with risk of permanency dissolution, the intersection of these demographic attributes may also matter. Specifically, if research on race and gender bias in evaluation of threat and perception of aggression (e.g., Halberstadt et al., 2018) extends to adoptive parents/guardians, then racial minority males may be at greater risk of permanency dissolution than white males or racial minority females, Similarly, age may exacerbate any racial bias in whether caregivers feel threatened by a child’s behavior, given evidence that Black youth, particularly males, are perceived as older, larger, and more physically imposing than White youth (Goff et al., 2014; Wilson et al., 2017). Prior literature has illustrated child behavior problems, mental health issues, or disability status consistently undermine permanency (Coakley & Berrick, 2008; Liao, 2016; Parolini et al., 2018), and therefore, are important to account for when examining direct and indirect associations between child characteristics and dissolution. The third aim of the current study was to investigate whether and how race moderated the association between other demographic characteristics and permanency dissolutions.

The Current Study

Given efforts to support adoption and guardianship as permanency options for children in foster care, it is important to explore their stability. Studies of the prevalence and risk factors for permanency dissolution typically focused solely on reunifications; those focused on adoption tend to involve small samples and were conducted between the 1970s to 1990s (e.g., Barth & Berry, 1988). Large-scale investigations of adoption and guardianship dissolutions are sorely needed. Thus, we investigated three research questions: (1) What is the rate of guardianship/adoption dissolution among a complete entry cohort of foster children in a large state?; (2) How do child demographic characteristics (race, age, and gender) influence the risk of guardianship/adoption dissolution?; and (3) Does child race moderate the association between other demographic characteristics and the risk of guardianship/adoption dissolution?

Method

Data

This study used data on a complete entry cohort of children who entered foster care in Texas during the 2008 and 2009 fiscal years. The Texas Department of Family and Protective Services (DFPS) extracted data from their administrative databases and provided de-identified and restricted data. This included all foster care placements from the point of entry until May 2016, which resulted in a cohort of 26,113 children. Children in the Other Race category were excluded from analyses due to small cell sizes with adoption and guardianship dissolutions and due to the ambiguous nature of this racial category. We restricted the sample to children who exited foster care to either adoption (N=7,538) or guardianship (N=4,950). Note: We use the common-use term guardianship throughout, but these arrangements are referred to as custody to a relative or permanent conservatorship in Texas.

Measures

Permanency was coded as adoption or guardianship based on a triangulation of discharge reason, caseworker reported placement end, and exit type. Dissolution was coded as an entry into foster care that followed a legal discharge to adoption or guardianship.

Race was categorized as Black, White, or Hispanic. Gender was coded as 1= male and 0=female. Age was based on the age in years of the child when they exited care. Risk factors were based on caseworker report at foster care entry: child behavior problem, mental health problem, cognitive or learning disability, and physical disability. In the adoption models, we also controlled for whether the adoptive parent was a relative of the child.

Covariates in our model included number of children in case, removal reason, number of placements, and duration. Number of children was a count of how many children were in the case. Removal reason included non-mutually exclusive dichotomous indicators of whether the child was removed due to: sexual abuse, physical abuse, emotional abuse, physical neglect, or need of supervision/caregiver substance abuse or mental health problem, and/or their caregiver was unable to provide care or child was in need of services. These variables were based on caseworker-reported reasons that children entered foster care. Number of placements included the total number of placements children experienced before adoption/guardianship placement. Duration was measured as the total time in care before adoption/guardianship placement.

Analytic Approach

We used multilevel survival models to model adoption and guardianship dissolution. Survival analysis is appropriate because our data are censored, in which the post-permanency observation period varied across children and many children remained at risk of, but did not experience, dissolution at the end of our observation period. Survival analyses allowed us to identify the risk of dissolution within a given time frame among those at risk during that period. The risk of dissolution began when the child exited foster care to a permanent setting and ended when any of the following occurred: a child reentered foster care, reached age 18, or met the end of the observation period (May 2016). Given the day-to-day operations of child welfare systems, and the family court system, are typically organized at the county level and the types of cases and practices can vary by county, our analyses included a random intercept for the child’s removal county. The random intercept model adjusts for the interdependency of observations within county. Dissolution was modeled separately for adoptions and guardianships. We estimated different hazard distributions to identify the best fitting model, and the AIC indicated that a Weibull distribution fit the data best for both adoption and guardianship dissolutions.

We note there is continued debate about the use of inferential statistics in administrative data (Gibbs, Shafer, & Miles, 2017). Given that our data include all individuals in Texas who entered foster care in fiscal years 2008–2009 and exited to adoption or guardianship, some may argue that our data constitute a population rather than a sample. Given our interest in applying our findings beyond a particular time frame and state, we characterize our data as a sample and report traditional measures of statistical significance. However, where potentially of practical importance, we also discuss patterns in coefficients that are marginally significant at p<.10.

Results

Descriptive Characteristics

A description of the sample by permanency type (adoption or guardianship) and dissolution status are found in Table 1. A little more than 2% of adoptive placements and 7% of guardianship placements ended with children returning to foster care. Adoptive placements were observed on average for 62.4 months after exiting care and guardianships were observed for an average of 70.4 months (range 0–104 months). Of those that dissolved, average time to dissolution was 33.1 months for adoptions and 27.8 months for guardianships. For both children in adoptive or guardianship placement, dissolution was more common among children with behavior problems, a cognitive disability, or mental health issues, and among children with more siblings and more pre-permanency foster care placements. Among adopted children, older age at adoption was strongly associated with dissolution. Time in care prior to permanency was negatively associated with guardianship, but not adoption, dissolution.

Table 1.

Descriptive Statistics by Dissolution Status

	Adoptive Placement (n = 7,538)			Guardianship Placement (n = 4,950)
	No Dissolution	Dissolution	Significant group differences	No Dissolution	Dissolution	Significant group differences
Demographic characteristics
Black	0.25	0.23		0.29	0.33
White	0.31	0.27		0.33	0.31
Hispanic	0.44	0.49		0.39	0.36
Male	0.50	0.50		0.51	0.47
Less than 2 years old	0.15	0.06	***	0.22	0.18	+
2 to 5 years old	0.45	0.31	***	0.35	0.37
6 years or older	0.40	0.63	***	0.43	0.45
Risk factors
Behavior problems	0.11	0.23	***	0.08	0.17	***
Cognitive disability	0.12	0.21	***	0.08	0.18	***
Mental health issues	0.04	0.09	***	0.03	0.09	***
Physical disability	0.05	0.04		0.03	0.04	+
Adoption type
Adopted by kin	0.46	0.44		-	-
Adopted by non-kin	0.54	0.56		-	-
Controls
Number of children on child’s foster care case	2.67 (1.59)	3.16 (1.40)	***	2.56 (1.42)	2.70 (1.40)	+
Removal reason
Sexual abuse	0.05	0.04		0.07	0.10	*
Physical abuse	0.35	0.27	*	0.32	0.30
Emotional abuse	0.02	0.01		0.03	0.03
Physical neglect	0.37	0.43		0.31	0.32
Supervision-related neglect	0.88	0.89		0.86	0.84
Caregiver unable to provide care or child is in need of services	0.33	0.34		0.31	0.37	*
Number of pre-permanency placements	2.73 (1.87)	3.31 (2.40)	***	2.50 (1.68)	2.98 (1.98)	***
Months in foster care before permanent placement	28.30 (13.90)	28.63 (11.93)		14.1 (10.50)	11.73 (7.65)	***

Note. Significance of group differences based on tests of equal means or proportions, as appropriate.

⁺p < .10

^*p<.05

^***p < .001.

Hazard Model Results for Adoption Dissolution

Multilevel survival models of the hazard of adoption dissolution are found in Table 2. Coefficients are reported as hazard ratios (HRs), which are interpreted similarly to odds ratios, where coefficient values greater than 1 indicate that a unit change in the independent variable is associated with an increased hazard of dissolution, and values less than 1 indicate that a unit change in the independent variable is associated with a decrease in hazard of dissolution. Model 1 is a basic model, and Models 2 through 4 add interaction terms to understand whether race is associated with dissolution for different subgroups. Across all models, the overall risk was low and decreased over time (Table 2). Associations between race and adoption dissolution were not statistically significant (see Model 1). Age at adoption was consistently and positively associated with adoption dissolution. Compared with children who were younger than 2 years of age of exit, the hazard of re-entering foster care was twice as high for children who were between 2 and 5 years old and nearly 5 times higher among children aged 6 years or older. Hazard of adoption dissolution was also substantially higher among children with a documented behavior problem or cognitive disability, but there were no significant differences in adoption dissolution based on documented mental health problems or physical disability. There were also no differences in adoption dissolution for children adopted by kin versus non-kin.

Table 2.

Risk of adoption dissolution (N= 7,536)

	Model 1 HR (SE)	Model 2 HR (SE)	Model 3 HR (SE)	Model 4 HR (SE)
Race (Reference = Black)
White	0.87 (0.20)	0.77 (0.28)	0.42 (0.32)	1.52 (0.51)
Hispanic	1.04 (0.22)	1.56 (0.46)	0.36 (0.27)	1.26 (0.40)
Gender (Reference = female)
Male	0.97 (0.15)	1.39 (0.45)	0.98 (0.15)	0.97 (0.15)
Age at Adoption (Reference = 0 to 2)
2 to 5 years old	2.11 (0.76)*	2.09 (0.76)*	1.28 (0.73)	2.08 (0.75)*
6 years or older	4.84 (1.81)***	4.71 (1.76)***	2.16 (1.23)	4.8 (1.79)***
Risk factors
Behavior problems	1.71 (0.37)*	1.70 (0.37)*	1.71 (0.37)*	1.72 (0.38)*
Cognitive disability	1.79 (0.36)**	1.81 (0.37)**	1.81 (0.37)**	1.79 (0.36)**
Mental health issues	1.32 (0.39)	1.35 (0.40)	1.32 (0.39)	1.30 (0.38)
Physical disability	1.02 (0.41)	1.02 (0.41)	1.02 (0.41)	1.00 (0.40)
Adoption Type (Reference = Kin adoption)
Adopted by non-kin	1.17 (0.19)	1.16 (0.19)	1.17 (0.19)	1.84 (0.61)+
Interaction Terms
Male * White		1.21 (0.53)
Male * Hispanic		0.44 (0.17)*
2 to 5 years old * White			1.55 (1.33)
2 to 5 years old * Hispanic			2.56 (2.13)
6 years or older * White			2.68 (2.19)
6 years or older * Hispanic			3.53 (2.85)
Adopted to non-kin * White				0.37 (0.16)*
Adopted to non-kin * Hispanic				0.70 (0.28)
Constant	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)
ln(p) [shape parameter]	0.08 (0.07)	0.08 (0.07)	0.08 (0.07)	0.08 (0.07)
Removal county variance	0.70 (0.31)	0.70 (0.31)	0.70 (0.31)	0.67 (0.30)

Note. Hazard Ratios (HRs, or exponentiated coefficients) are presented. SE=standard error. Reference group for age at exit is younger than 2 years old. Reference group for race is Black. Reference group for control variables is adopted to kin.

⁺p < .10

^* p < .05

^** p < .01

^***p < .001.

All models controlled for number of children in case, removal reason(s), number of placements, and total time in care.

In the next set of models, we explored whether race moderated the association between other demographic characteristics and adoption dissolution. In an interaction model, the interpretation of the coefficient for the interacted variables changes and should not be directly compared to the coefficients in the non-interacted models. Thus, the predicted hazard has been plotted for each of the significant interaction terms for ease of interpretation (Figures 1 and 2). Model 2 (of Table 2) included a child gender * race interaction and revealed that Hispanic females had a higher hazard of adoption dissolution than White females and Hispanic males (p <.05; see Figure 1). Further, although pairwise comparisons were not statistically significant at traditional levels, Black and White males may be at higher risk of adoption dissolution than White females (p <.10). Model 3 included a race * age interaction term at foster care exit interaction, but this were not statistically significant.

Figure 1.

Predicted hazards of adoption dissolution by race and gender.

Figure 2.

Predicted hazards of adoption dissolution by race and adoption type (kin or non-kin).

Lastly, Model 4 included a race * adoption type (kin or non-kin) interaction, which indicated White-Black differences in risk of non-kin adoption dissolution. Post-hoc testing revealed that none of the group comparisons differed at p<.05. Of practical importance; among Black children, those adopted by non-kin had a higher hazard of adoption dissolution than those adopted by kin, and White children had a lower risk of adoption dissolution than Black or Hispanic children in non-kin placements (p<.10; see Figure 2).

Hazard Model Results for Guardianship Dissolution

Results for guardianship dissolution are shown in Table 3. The initial risk of guardianship dissolution was low and decreased over time. Model 1 indicates that White children (HR = .76, p < .10) and Hispanic children (HR = .72, p < .05) had lower hazard of guardianship dissolution than Black children. Male children had a lower hazard of foster care reentry than females (HR = .80, p < .05). Lastly, children who exited to guardianship between 2 to 5 years of age had a higher hazard of dissolution than children who exited care before 2 years old. Children with a behavior problem, cognitive disability, or mental health problem had nearly double the hazard of experiencing a guardianship dissolution compared to children without any risk factors. Results from Models 2 and 3, in Table 3, suggest that race does not moderate the associations between gender nor age and guardianship dissolution.

Table 3.

Risk of guardianship dissolution (N= 4,947)

	Model 1 HR (SE)	Model 2 HR (SE)	Model 3 HR (SE)
Race (Reference = Black)
White	0.76 (0.11)+	0.81 (0.16)	0.67 (0.22)
Hispanic	0.72 (0.11)*	0.79 (0.15)	0.82 (0.24)
Gender (Reference = female)
Male	0.80 (0.09)*	0.89 (0.17)	0.80 (0.09)*
Age at Adoption (Reference = 0 to 2)
2 to 5 years old	1.35 (0.21)+	1.35 (0.22)+	1.19 (0.33)
6 years or older	1.16 (0.20)	1.16 (0.20)	1.32 (0.35)
Risk factors
Behavior problems	1.80 (0.33)**	1.81 (0.33)**	1.78 (0.33)**
Cognitive disability	2.03 (0.31)***	2.02 (0.31)***	2.03 (0.31)***
Mental health issues	1.91 (0.46)**	1.89 (0.45)**	1.91 (0.46)**
Physical disability	1.25 (0.34)	1.25 (0.34)	1.22 (0.34)
Interaction Terms
Male * White		0.88 (0.23)
Male * Hispanic		0.83 (0.21)
2 to 5 years old * White			1.36 (0.54)
2 to 5 years old * Hispanic			1.10 (0.40)
6 years or older * White			1.03 (0.40)
6 years or older * Hispanic			0.69 (0.24)
Constant	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)
ln(p) [shape parameter]	−0.25 (0.05)	−0.26 (0.05)	−0.25 (0.05)
Removal county variance	0.88 (0.26)	0.88 (0.26)	0.88 (0.26)

Note. Hazard Ratios (HRs, or exponentiated coefficients) are presented. SE=standard error. Reference group for age at exit is younger than 2 years old. Reference group for race is Black. Reference group for control variables is adopted to kin.

⁺p < .10

^*p < .05

^**p < .01

^***p < .001.

All models controlled for number of children in case, removal reason(s), number of placements, and total time in care.

Discussion

This study examined prevalence of adoption and guardianship dissolution and how demographic characteristics and risk factors predicted dissolution. There is currently limited research on the frequency and predictors of adoption dissolution and even less research on guardianship dissolution. To address this gap in the literature, this study used administrative data from a complete entry cohort of children who entered foster care in a large southern state.

The overall rate of dissolution was higher for guardianship placements (7%) than adoptions (2%), which are similar to rates in prior work on adoption or guardianship dissolutions (Coakley & Berrick, 2008; Goering & Shaw, 2017; Parolini et al., 2018; Rolock, 2015; Rolock & White, 2017). However, both adoption and guardianship dissolution were rare when compared with rates of reentry following reunification, which typically range from 20–40% (Kimberlin et al., 2009) and were estimated at 18% in Texas (Font et al., 2018).

Of note, this was one of the first studies to compare adoption and guardianship dissolution rates in the same sample. These results add to existing evidence that guardianships are more precarious than adoptions. One possibility for this difference in rates could be differences in caregiver motivations. Adoptive parents are often motivated by a desire to provide a permanent home for a child and a desire to start or expand a family (Malm & Welti, 2010). Guardians, who in this sample are nearly all relatives of the child, typically assume custody because of a family event. Guardians may be motivated by familial obligation and a sense of duty to a relative child, as prior research on kinship foster parents suggests (Testa & Slack, 2002). As such, some guardians may agree to care for a child without fully considering the potential economic or interpersonal hardships. Further, although financial subsidies are available for guardianships, similar to those available for adoption, relatively few guardianships in Texas received subsidies because the guardians were not licensed foster parents. Thus, financial hardship may also contribute to higher dissolution rates for guardianships than adoptions. Another possibility is there may be differences in the characteristics of caregivers that pick guardianship over adoption. Previous research has shown that certain caregiver characteristics are related to a higher risk of dissolution, such as younger age, lower levels of education, lower income, and unmarried caregivers (Coakley & Berrick, 2008; Liao & White, 2014; Testa et al., 2015). Although previous work has shown that children in kinship care are more likely to experience guardianship than adoption and kin caregivers tend to prefer guardianship over adoption (see review Bell & Romano, 2015), there is a limited understanding of whether certain caregiver characteristics among kin predict adoption or guardianship and whether those same characteristics undermine permanence. Further research on the preparation, motivation, and characteristics of guardians and systematic tracking of adoption and guardianship dissolution rates are needed to understand why guardianships dissolve at a higher rate than adoptions and how policy and practice can be adapted to lower dissolution rates.

Consistent with prior research (Parolini et al., 2018; Rolock & White, 2016), our study further suggests that both White children and Hispanic children are less likely to experience guardianship dissolution than Black children. Guardianship has been promoted as a permanency alternative to adoption without termination of parental rights (Rowe, 2013), yet our results question whether guardianship and adoption are equally stable environments. If Black children were more likely to exit to guardianship, this could create a disparity in reentry rates. Akin (2011) found that Black children, compared to White children, were significantly less likely to exit to adoption, but equally likely to exit to guardianship. In other words, Black children are more likely to remain in foster care in the first place or return to foster care following guardianship placements. In contrast to previous research findings that racial minority children experienced higher rates of adoption dissolution (Orsi, 2015; Smith et al., 2006), we found that race was not associated with adoption dissolution. Future work should continue to explore whether and how race is associated with risk of adoption and guardianship dissolution.

Other demographic characteristics were also important predictors of permanency dissolutions, but these associations depended on whether it was adoption or guardianship. Age was an important predictor of adoption dissolution, consistent with prior research (e.g. Coakley & Berrick, 2008). The adoption of older children in foster care has been a longstanding federal priority, with financial incentives provided to states to promote the adoption of children 9 years or older (U.S. Congressional Research Service, 2013). Although recent research suggests these incentives were not effective, such efforts persist under the expectation that adoption is beneficial for all children (Brehm, 2018). Yet, there is little reason to believe that adoptions that dissolve are preferable to remaining in long-term foster care. Nevertheless, states are still expected to find permanent placements for children at high risk for dissolution. Our study highlights that finding an adoptive home for an older child in foster care is the first step to securing permanency, not the last. Current financial subsidies for adoptive families may be beneficial, but access to high-quality post-permanency services may also be necessary. Families frequently need counseling/psychological services, medical services, housing services, assistance with child care, and educational services (Liao & White, 2014; Armstrong, Johnson, Robst, Cruz, Landers, & Vargo, 2019). Yet, families report they are sometimes unable to access needed services or face substantial delays in access (Liao & White, 2014; Armstrong et al., 2019). Even when services are received, both families and staff have questioned their quality and effectiveness (Armstrong et al., 2019). Thus, providing high-quality services in a timely manner might be most beneficial to adoptive and guardianship placements.

For children in guardianships, males were less likely to experience guardianship dissolution, which has some support from previous work (Parolini et al., 2018). Although we cannot elucidate the underlying mechanisms on why females were at higher risk, we consider one possibility. Male children in our sample were more likely to be identified as having behavior problems or cognitive disability. It is possible that behavior problems in females are less likely to be documented than behavior problems in males, even if they exhibit similar behaviors. Thus, caregivers of female children may be expecting minor or nonclinical levels of behavior problems and be disillusioned by the challenges of the placement. Dissolutions, like other placement disruptions, may reflect a “mismatch” between caregivers’ skills and expectations and a child’s needs and risk behaviors (Sattler et al., 2018). The availability of accurate and complete information about the child is important when prospective guardians are deciding whether they can meet the needs of the child and when agencies are determining how much support to provide families. Although prospective guardians are typically relatives and may even be a foster parent to the child prior to assuming guardianship, this does not guarantee they have all relevant information about the needs or health conditions of the child.

Caseworker reports of child behavior problems and cognitive disability status were associated with higher risks of adoption and guardianship dissolution, and a mental health problem was related to a higher risk of a guardianship dissolution. Prior evidence has shown behavioral, cognitive, and emotional problems all being related to a higher likelihood of adoption dissolution (Coakley & Berrick, 2008; Liao, 2016; Parolini et al., 2018). We caution that most children who enter foster care have experienced adversities, such as child abuse and neglect, and the rates of mental and behavioral health challenges likely exceed those reported in the data. Still, our findings suggest these indicators may capture more severe behavior, mental health, or cognitive challenges that may be difficult for adoptive parents and guardians.

Our final aim was to explore whether race moderated associations of gender, age, and adoption type with permanency dissolution. We found no evidence of moderation effects for guardianship dissolution. In contrast, race moderated associations of gender and adoption type (kin versus non-kin) with adoption dissolution. White females had the lowest risk of adoption dissolution, whereas Hispanic females and Black males had elevated risk. These results show a complicated relationship between race, gender, and adoption dissolution that requires further research. Although our bivariate statistics indicated that kin and non-kin adoptions were equally unlikely to dissolve, our multivariate models indicate this finding does not hold for Black children. Black children adopted by non-kin had a higher risk of adoption dissolution than Black children adopted by kin, whereas either adoption type was equally stable for White children. Among non-kin adoptions, Black and Hispanic children had a higher risk of adoption dissolution than White children. These findings illustrate potential vulnerabilities in non-kin adoptions for racial and ethnic minorities. Our study was not able to specifically assess whether adoptions were transracial, but that is an important area for future research. It may be children’s demographic characteristics should be considered when determining the training, resources, or supports an adoptive family may need to be successful.

Several limitations should be considered when evaluating these results. First, our measure of dissolution only captures children who reenter foster care. Some guardianships or adoptions may end without a child reentering foster care (e.g., if a different relative assumes custody). Further, this paper did not capture children that remained in care at the end of the observation period; however, this was approximately only 2% of the sample. Second, even with a large total sample, our statistical power was limited due to low rates of dissolution. This resulted in imprecise estimates for some coefficients and limited the number of covariates we included in our models. With adequate power, we would have included CPS history (no prior removals, prior allegation no removal, or prior removal), level of care (any intensive/psychiatric or not), and any time spent in restrictive care. Each of these variables would likely capture children with more severe behavior problems/mental health issues and could potentially lessen some of the current associations. Third, this study is descriptive in nature and more research is needed to uncover the mechanisms that explain differences in adoption and guardianship dissolution. The administrative data used for this study lacked rich information on child and family attributes, which are needed to illuminate potential confounding variables or mechanisms. There are likely unobserved factors that confound associations between children’s characteristics and permanency dissolution, such as the quality of adoptive or guardianship home or caregiver’s socioeconomic status. Future work could use survey data or augment administrative data to account for a broader range of variables. Lastly, reports of children’s disability status and behavioral problems underestimate the behavioral and mental health challenges of children exiting foster care and provide little context on the severity or nature of disability or behaviors. Higher quality data on children’s health and well-being in foster care is needed.

Nevertheless, this paper has several strengths, including a large, longitudinal, and diverse sample, separate analysis of adoption or guardianship dissolutions, and analyses that adjusts for county-level clustering and censoring of data. Although dissolution is a relatively uncommon event, its potentially devastating impacts on children warrant attention from researchers. When children reenter foster care, they may be reluctant to build new relationships and have difficulty trusting others. In addition, prospective adoptive or guardianship parents may view a prior dissolution as a serious red flag about the child’s behavior or ability to form attachments, thus increasing the probability that children who experience dissolution will “age out” of the system. Our findings suggest that increased attention to the stability of guardianships and non-kin adoptive placements of racial and ethnic minorities is necessary. It is not always clear when guardianship is a suitable or preferable alternative to adoption. When caregivers express interest in guardianship but not adoption, it may be appropriate to assess whether such preferences are driven by hesitance about providing a permanent home, and to ascertain whether guardianship reflects the desires and best interests of the child. In addition, to reduce the heightened risk of guardianship dissolution, access to guardianship subsidies and ongoing Medicaid coverage should be provided to all children exiting foster care to guardianship, regardless of Title IV-E eligibility or caregiver income. Kinship Navigator programs are also a promising practice for supporting safe and stable kinship care, including guardianships (Child Welfare Information Gateway, 2018) and the expansion of these programs following the Family First Prevention Services Act (2018) may facilitate improved service quality and access for guardians. Lastly, our results provide evidence that some children, particularly Black, older, or female children or children with behavior or mental health needs, may be more vulnerable to experiencing dissolution. As such, certain permanency placements may need additional services or supports to promote long-term stability. Policymakers may also need to identify the best ways to counteract inequalities that differentially impact specific demographic groups.